Faecal indices predict organic matter digestibility, short chain fatty acid production and metabolizable energy content of browse-containing sheep diets

The ability of faecal composition to accurately predict nutritional characteristics of browse-containing sheep diets was investigated. Predictive regression models were developed using data derived from three feeding experiments conducted with Acacia saligna (i.e., Avena sativa control and five inclusion levels of A. saligna), Chamaecytisus palmensis (i.e., A. sativa control and five inclusion levels) and Atriplex amnicola (i.e., A. sativa control and five inclusion levels). The fourth experiment conducted to validate the predictive models included six levels of A. saligna and five levels of C. palmensis. Feed and faecal samples collected from six replicate sheep were analysed for neutral detergent fiber (NDFom), acid detergent fiber (ADFom), lignin(sa), ash and crude protein (CP) contents. Total phenolics (TP) and tannin (TT) contents and in vitro gas production of the feed samples were also measured. Organic matter digestibility (OMD), short chain fatty acid production (SCFA) and metabolizable energy (ME) content of diets were determined. Feed and faecal data of experiments 1, 2 and 3 were pooled. Correlation coefficients among feed and faecal variables were estimated. The predictive regression models of dietary characteristics were developed from faecal composition by a stepwise regression procedure. The significant (i.e., P<0.05) predictive model with the highest R2 and lowest residual standard deviation (RSD) was selected (i.e., best-fit predictive model). The model was used to predict nutritional characteristics of diets in the validation experiment from respective faecal indices. The predicted nutritional characteristics were then regressed against measured nutritional characteristics. The best-fit predictive regression model for dietary CP had a very low R2 (0.21). The best-fit regression models predicting dietary TP, TT, OMD, SCFA and ME from faecal N, ash and lignin(sa)/NDFom had R2 greater than 0.78 and low RSD. In the validation experiment, regressions between measured and predicted TP and TT had positive intercepts (P<0.05) and low R2. Slopes of these regressions were much lower than 1.0. Intercepts of the regressions between measured and predicted OMD, SCFA and ME did not different from zero. The R2 of these regressions were very high and slopes were close to 1. Dietary CP content of browse-containing diets cannot be predicted from faecal fiber, lignin(sa) and ash contents. However, these faecal indices collectively can predict the OMD, SCFA production and ME contents of browse-containing diets, and these predictive models have broad application. Further research is necessary to understand the poor validity of the predictive models of TP and TT.